Skip to content

pullrequests Search Results · bucket language:JavaScript language:JavaScript language:Python language:JavaScript language:Python

Filter by

916k results  (896 ms)

916k results

dependencies
patch

Summary - compile SM90 sparse indexer-score and sparse attention-score kernels during plan preparation from symbolic tensors - bound specialization to 17 physical Top-K capacity buckets while ...
area:sparse_attention
cat-feature
op: DSA
orig-external

What does this PR do? In Bot Mode a bot s chat is a session tile and the primary selection is empty. The preview rail picks its profile bucket from the primary runtime (railScopeForActiveSession ...
area/profiles
area/sessions
comp/desktop
P2
type/bug

... the batch-size buckets as: max_bs = min(self.config.max_num_seqs, 512) self.graph_bs = [1, 2, 4, 8] + list(range(16, max_bs + 1, 16)) which leaves out max_bs itself unless it is a multiple of 16 ...